The present disclosure relates generally to data warehousing, analytics and appliances and more particularly to Massively Parallel Processing (MPP) database backup/restore/migration optimization using directed backup.
An MPP database partitions data across multiple servers, or nodes, with each node having its own processor and memory to process data locally. Each node server controls a portion of the database and a portion of the database storage. All communication is via an “interconnect” as there is no disk-level sharing. There is also no contention to the data since it is a ‘shared-nothing’ architecture.
An MPP database may be controlled by a “host” processor with its own operating system, storage and memory. The host system does not necessarily store user data, but rather partitions the user data across multiple nodes to achieve data and processing parallelism. Typically, an MPP host processor partitions a database among the nodes and assigns work among the node servers.